Demand Sensing, a Textbook Illustration of Mootware

Demand Sensing is a forecasting method that combines the latest mathematical techniques and real-time information to create an accurate forecast of demand. It is also yet another buzzword in the supply chain industry.

In this episode of LokadTV, we discuss how well it works and learn what a supply chain practitioner can do to sort the good buzzwords from the bad to be able to see what in fact is simply a marketing gimmick.

The basic concepts advertised by Demand Sensing vendors are a use of real-time data and machine learning, but most of what is promised by Demand Sensing documentation is somewhat vague and undefined. Overall, Demand Sensing could be described not as vapourware (a software or hardware that exists but is not yet available to purchase, as it is still in the concept or design stage), but as “mootware”, i.e. a software that won’t deliver what it claims.

The various ideas behind Demand Sensing stand up on their own, but why does Demand Sensing as a whole seem so unrealistic? For example, how can you qualify real time data, which is one of the key elements proposed by Demand Sensing? It could be said that real time data is when the latency goes below human perception, for example 100 milliseconds. This isn’t strictly below human perception but it’s very close to what we can qualify as “real time”.

In a supply chain context, let’s imagine we’re forecasting demand for a product that has three years worth of sales history for the sixth months ahead. Here, does having “real time data” fresh from the past 100 milliseconds when compared to data from the past 24 hours or two days ago actually make any considerable difference to the forecasting accuracy?

Along with this real time data aspect, Demand Sensing promises “superior” machine learning techniques. Yet machine learning is an extremely broad domain, so this is a very vague claim with little substance or clarity provided by any of the Demand Sensing documentation.

Shallow buzzwords will always be an issue in the tech industry - for example, IBM pushing their “autonomous computing” ten years ago. To wrap up this episode, we discuss concrete ways in which you can see if concepts linked with buzzwords have any substance to them and if they’re worth pursuing for your business.

Timestamps

00:08 Introduction

00:29 What is your initial overview of the concept?

02:44 What indicates it is a marketing gimmick?

04:43 What are the problems?

08:35 Which level of granularity do we need?

09:46 Who is it benefitting then?

11:24 How do we differentiate between what is a good buzzword and what is not?

14:22 Are there lots of other examples of gimmicks that have not done so well historically?

16:18 What is the outcome?

17:35 What should supply chain practitioners be on the lookout for?